MKTG 443: Digital Marketing & Social Media - AI for Digital Marketing Notes

Course Overview and Administrative Details

• Course Code and Name: MKTG 443 - Digital Marketing & Social Media.

• Semester: Fall 2025.

• Instructor: Luxi Chai, Ph.D.

• Institution: KU School of Business, The University of Kansas.

• Website: business.ku.edu.

Upcoming Schedule and Activities

• April 21 (4/21): Coverage of Chapter 12: Mobile Marketing; Work on Personal Website.

• April 23 (4/23): Coverage of Chapter 14: AI for Digital Marketing; Case 2 study.     * Deadlines: Quiz 5 is due; Personal Website is due.

• April 28 (4/28): Case 2 review; Exam Review session.

• April 30 (4/30): Exam 2.     * Deadlines: Google Analytics 4 (GA4) certification submission due (reference sample submission on Canvas); Additional Certificate due.

• May 5 – May 7 (5/5-5/7): Pharm D project activity.     * Deadlines: Part 2 of the project is due.

Google Analytics 4 Certification Reference

• Completion Sample 1 (Chance White):     * Date of Completion: November 8, 2024.     * Course: Get started using Google Analytics.     * Certificate ID: $57183121858192$.     * Final Score: $96\%$.

• Certification Sample 2 (Jillian Alston):     * Issue Date: November 18, 2024.     * Expiry Date: November 18, 2025.     * Certificate ID: $123187239$.     * Title: Google Analytics Certification.

Fundamental Artificial Intelligence (AI) Concepts

• Artificial Intelligence (AI): Defined as the broad umbrella of technology where machines perform tasks that normally require human intelligence. Examples include chatbots, recommendation systems, and self-driving cars.

• Machine Learning (ML): A subset of AI focused on teaching machines how to perform specific tasks and provide accurate results by identifying patterns from data. Examples include predicting sales, spam detection, and product recommendations.

• Deep Learning (DL): A subset of ML that focuses on developing artificial neural networks capable of learning from massive amounts of data. Examples include image recognition, speech-to-text, and language translation.

• Artificial General Intelligence (AGI): A theoretical computer that can mimic the problem-solving and decision-making abilities of humans.     * Note: Standard machine learning does not reach the level of AGI.     * Status: Modern tools like ChatGPT suggest AGI may be within reach. It is evolving so rapidly that it is difficult to predict even the near future.

• Neural Network: The primary tool used in Deep Learning; it is a model inspired by the human brain consisting of interconnected "neurons." These power image classifiers and voice assistants.

• Large Language Model (LLM): A massive neural network trained specifically to understand and generate human language.     * Pattern Extraction: LLMs enable neural networks to extract meaning from word patterns and generate meaningful writing.     * Examples: ChatGPT, GPT-5, Claude, Gemini.     * Hallucination: A phenomenon where the AI produces incorrect but plausible-seeming statements (Fits ≠ Truth).

• Computer Vision (CV): An area of computer science enabling computers to "see" and understand image or video content. It utilizes Deep Learning for tasks like Face ID and medical image analysis.     * MNIST Database Example: A standard for CV where each digit is a $28 \times 28$-pixel grid of values.     * Values: Ranging from $0$ to $1$, where $0$ represents white and $1$ represents black.     * Total Data Points: Each image contains $784$ values ($28\times28$).

Mathematical Framework of AI Operations

• The Model Equation: The basic operation of an artificial neuron is represented by the formula:     $\hat{y} = g(w \cdot x + b)$

• Variables Defined:     * $x_1, x_2, … x_n$: The Inputs.     * $w_1, w_2, … w_n$: The Weights applied to inputs.     * $b$: The Bias.     * $\Sigma$: The Summation of weighted inputs and bias.     * $g(.)$: The Activation function.     * $\hat{y}$: The predicted Output.

Categorization of AI Types

• Classic AI:     * Function: Identifies patterns and predicts outcomes.     * Marketing Example: Identifying which specific zip code has the highest concentration of potential buyers.

• Interactive AI:     * Function: Converses and takes actions.     * Marketing Example: Assisting a customer in tracking their order through an automated interface.

• Generative AI:     * Function: Creates and imagines new content.     * Marketing Example: Drafting an original blog post for a new product launch.

AI in Digital Advertising and Campaign Optimization

• Prerequisites for AI Advantage:     * Algorithms require massive amounts of data to function effectively.     * Algorithms are limited to optimizing exactly what they are instructed to optimize (the objective function).     * AI-driven optimization can make attribution numbers more misleading if not analyzed carefully.

• Ad Creation Utility:     * Value Proposition: AI is used to create and refine ad copy to present the best possible value proposition. Effectiveness in communicating this value is the primary determinant of ad success.     * Targeting: AI matches specific value propositions with relevant ad targets.

• Role of Generative AI: Extremely helpful in brainstorming ideas for search and display ads. The quality of output is strictly tied to the "Quality of the Prompt."

Prompt Engineering Exercises and Examples

• Carpet Cleaning Case Study ("Deep Clean"):     * Standard Prompt: "Create a text-only search ad for a carpet cleaning service called 'Deep Clean'."         * Generated Elements: Taglines like "Spotless Carpets Guaranteed!", bullet points on eco-friendly technology, safe for pets/children, and a placeholder for contact details.     * Specific Prompt A: Focusing the ad specifically on a price-match guarantee and a money-back guarantee.     * Specific Prompt B: Focusing the ad on an "exclusive chemical cleaning process" that removes stains that competitors cannot.

• Protein Bar Exercise (Prompt Depth Comparison):     * Step 2a (Simple): "write an ad copy of a protein bar" → Results in a generic ad.     * Step 2b (Medium): "Write a short and catchy Instagram ad for a new high-protein, low-sugar energy bar for college students who need a quick, healthy snack." → Content becomes more targeted.     * Step 2c (High-Detail/Structured): "Write a persuasive Instagram ad ($40–50$ words) for a new high-protein, low-sugar bar called PowerBite. Target audience: busy college students. Tone: energetic, fun, Gen Z style. Include a call to action and one emoji." → Results in a much stronger, relevant ad.

Guidelines for Utilizing AI in Marketing

1. Include Segment Details: AI prompts must always include specific details about the customer segment being targeted.

2. Human Verification: Final approval of AI-generated responses should always be performed by humans, preferably a team of multiple people.

3. Varied Ideation: Marketers should explicitly ask the AI to produce highly varied ad ideas for each customer segment to avoid repetition.

AI Applications in Search and Content Management

• SEO (Search Engine Optimization):     * AI helps generate a list of informational pages on relevant topics to build topical authority.     * Process: AI creates the first draft; humans then thoroughly edit it to "humanize" the content.     * Velocity: Content should be added in a "steady stream" rather than all at once to maintain a reasonable timeframe.

• SEM (Search Engine Marketing / Paid Search):     * Topical Authority: Aligning content with surrounding context words.     * Search Generative Experience (SGE): The rise of "zero-click" searches where AI answers the query on the search result page.     * Smart Bidding: Automated bidding systems where attribution difficulties often arise.     * Role Shift: The role of a marketer is shifting from being a "copywriter" to a "director" of creative content.

Broader Marketing Applications of AI

• Marketing research and consumer insights.

• In-depth consumer analysis.

• Personalization of user experiences.

• Email marketing and social media management.

• Customer service automation.

AI Ethics, Law, and Misinformation

• Intellectual Property (IP): A response produced by ChatGPT (text or image) is considered original. It is not a direct copy of existing work and therefore generally does not violate copyright law or IP rights as currently structured.

• Pattern Bias:     * Generative engines are trained on existing documents and replicate patterns in those datasets.     * Result: Existing societal biases are replicated in output.     * Example: Prompting "Show a picture of a Fortune 100 CEO giving a speech" might result in biased demographic representations based on historical data patterns.

• Misinformation and Manipulation:     * AI can magnify malicious persuasion attempts.     * Targeting algorithms can be misused for election interference (e.g., persuading specific groups not to vote).     * Deepfakes: Realistic videos of public figures saying or doing things they never did. These can cause irreparable damage to reputations and the truth.